Morphological operations (erosion, dilation, opening, closing) in Computer Vision

Introduction

Morphological operations help clean and change shapes in images by adding or removing pixels. They make it easier to find important parts or fix small mistakes in pictures.

Removing small noise dots from a scanned document image.

Making objects in a photo thicker or thinner to see their shape better.

Separating two objects that are touching in a picture.

Filling small holes inside objects in a black and white image.

Smoothing the edges of shapes in a binary image for better analysis.

Syntax

Computer Vision

cv2.erode(image, kernel, iterations=1)
cv2.dilate(image, kernel, iterations=1)
cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel)
cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)

image: Input image, usually black and white (binary) or grayscale.

kernel: A small matrix (like 3x3) that defines the shape of the operation.

iterations: How many times to apply the operation.

Examples

This makes the white parts in the image smaller by removing pixels at the edges.

Computer Vision

kernel = np.ones((3,3), np.uint8)
eroded = cv2.erode(image, kernel, iterations=1)

This makes the white parts bigger by adding pixels around edges, done twice.

Computer Vision

kernel = np.ones((5,5), np.uint8)
dilated = cv2.dilate(image, kernel, iterations=2)

Opening removes small white noise by eroding then dilating.

Computer Vision

kernel = np.ones((3,3), np.uint8)
opened = cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel)

Closing fills small black holes by dilating then eroding.

Computer Vision

kernel = np.ones((3,3), np.uint8)
closed = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)

Sample Model

This code creates a small black and white image with a white square and one noise pixel. It then shows how erosion, dilation, opening, and closing change the image.

Computer Vision

import cv2
import numpy as np

# Create a simple binary image with noise
image = np.zeros((7,7), dtype=np.uint8)
image[2:5, 2:5] = 255  # a white square
image[1,1] = 255      # noise pixel

kernel = np.ones((3,3), np.uint8)

# Apply erosion
eroded = cv2.erode(image, kernel, iterations=1)

# Apply dilation
dilated = cv2.dilate(image, kernel, iterations=1)

# Apply opening (remove noise)
opened = cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel)

# Apply closing (fill holes)
closed = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)

print("Original image:\n", image)
print("Eroded image:\n", eroded)
print("Dilated image:\n", dilated)
print("Opened image:\n", opened)
print("Closed image:\n", closed)

Important Notes

Erosion shrinks white areas, useful to remove small white noise.

Dilation grows white areas, useful to fill small black holes.

Opening is erosion followed by dilation, good for removing noise without changing shape much.

Closing is dilation followed by erosion, good for filling holes inside objects.

Summary

Morphological operations change image shapes by adding or removing pixels.

Erosion shrinks white parts; dilation grows them.

Opening removes noise; closing fills holes.